The conversation around AI agents is changing. For many enterprises, the question is no longer whether to deploy them, but how to make multiple agents, employees and systems work together without creating a fragmented customer experience.
Genesys' announcements at Xperience 2026 offer a good example of where the market is heading. Rather than treating AI as another tool for customers or employees, Genesys is putting orchestration at the centre of its strategy, connecting what a customer wants with the systems and people needed to deliver it.
Orchestration at the centre of Genesys Cloud
At Xperience, Genesys introduced four capabilities that together form a broader orchestration architecture. Genesys Cloud Navigator interprets customer intent and determines the appropriate path or resource. Genesys Cloud Orchestrator coordinates AI agents, employees, systems and workflows to reach an outcome. Contextual Intelligence carries customer context across interactions and journeys, while the AI Control Plane provides governance, policy control, visibility and traceability as AI takes on more autonomy.
In Genesys' announcement, IDC research manager Hayley Sutherland described the suite as bringing together "context, orchestration and governance across the customer journey." Genesys Cloud Contextual Intelligence and the AI Control Plane are available now, while Navigator and Orchestrator are scheduled for later releases.
Pinkfish was the groundwork
The launch also makes more sense in the context of Genesys' acquisition of Pinkfish in June. The acquisition added enterprise workflow automation and connectivity to the platform, with more than 500 pre-built integrations and roughly 25,000 MCP tools covering systems such as CRM, ERP, billing, HR and order management.
When you put the two announcements together and a clearer picture emerges. Having already acquired the technology to connect AI to enterprise systems, Genesys is now putting that capability into a wider orchestration strategy.
AVA moves from conversation towards execution
Genesys has also expanded its Agentic Virtual Agent (AVA). What is interesting about this isn't just that AVA can have better conversations, but what the agent can do once the conversation is underway.
AVA now has access to Scaled Cognition's APT-2 action model, new voice capabilities from ElevenLabs and Deepgram, and, according to Genesys, deeper connectivity to the enterprise integrations and MCP tools available through the Pinkfish technology. It can also work with other AI agents through A2A, while MCP provides access to tools and enterprise systems.
Consider a customer reporting a delayed delivery. Instead of simply checking the order status and explaining the delay, an agent could access the relevant systems, take an authorised action and resolve the issue.
Genesys doesn't need to own every agent
Genesys is also leaning on its relationships with Salesforce and ServiceNow rather than trying to build a single universal agent itself. A separate announcement at Xperience confirmed new and expanded collaborations with Adobe, AWS, Deepgram, ElevenLabs, Meta, Salesforce, ServiceNow and Sierra, with A2A interoperability specifically letting AVA collaborate with agents from Salesforce Agentforce and ServiceNow.
In a Forbes interview published the same day, CEO Tony Bates spoke about its strategy: "We're not asking the customer to choose between their existing platforms. We're helping coordinate them around the outcome the customer is trying to achieve."
The bigger CX AI shift
Scaling agents changes the problem enterprises face. As AI agents move from pilots into production, the challenge for CX leaders is changing from making individual agents work well on their own to making multiple agents, across many systems, work well together. This makes orchestration, context and governance increasingly important.
The emerging competition in CX AI may, therefore, not simply be about which vendor has the smartest agent, but who controls the layer connecting customer intent to enterprise action. Genesys’ latest announcements demonstrate its view that the CX platform should own that layer.